238 | Scott Shapiro on the Technology and Philosophy of Hacking
What this covers
Scott Shapiro, a law professor and philosopher at Yale, sits down with Sean Carroll to argue that hacking is not a glitch to be engineered away but a structural inevitability built into the nature of computing itself. The conversation moves between the mathematical foundations of computation and the real-world history of major breaches—from the 1988 Morris worm to the DNC hack tied to Russian military intelligence—to make a singular point: unhackable computers are philosophically impossible. Shapiro traces this back to Turing's 1936 insight that code and data can be represented by the same symbols, which is precisely what attackers exploit. Because general-purpose computers rely on this equivalence, eliminating hacking would mean eliminating the feature that makes computers powerful.
Yet the episode's real force lies not in abstract argument but in how Shapiro moves from the mathematical to the human. Von Neumann's principle that self-replication makes biological systems resilient is the same principle that enables computer viruses; phishing works because a campaign manager's IT consultant types "this is a legitimate email" when meaning the opposite. From the default passwords on the Mirai botnet to the social psychology of young coders seeking peer recognition in underground forums, Shapiro argues that [upcode](jargon:upcode)—human norms, incentives, and policy—matters more than technical [downcode](jargon:downcode) fixes. He traces much cybercrime not to malice but to underemployment in less developed tech economies, suggests legitimate hacking competitions as a remedy, and shows how a single breach of international norms—Fancy Bear's public dump of stolen data—transformed espionage into what some call cyber war. The result is a case that cybersecurity is ultimately a problem for law and sociology as much as engineering.
Shapiro argues that hacking is not a fixable flaw but an inevitable consequence of the very principles that make general-purpose computing possible, and that the most effective defenses lie in 'upcode' (human norms, incentives, and policy) rather than purely technical 'downcode' fixes.
- Turing's insight that code and data can be represented by the same symbols (duality) is exactly what hackers exploit, so unhackable computers are impossible.
- Human beings are the weakest link, so cybersecurity must address psychological and institutional incentives, not just technical vulnerabilities.
- Most cybercrime traces to social conditions like underemployed young men seeking clout, suggesting social interventions are effective remedies.
Hackability is inherent to computation itself, not a flaw solvable by resources or effort.
- The principles Alan Turing laid out in 1936 for how general computing devices are possible turn out to be exactly the principles hackers exploit, meaning the very things that make computers possible make hacking possible.
“the principles that he lays out for how general computing devices are possible, which we can go into, it turns out to be exactly the principles that hackers exploit when they hack a computer”
- It is impossible to build unhackable computers—not because people are careless or make mistakes, but because hackability is part of the fundamental fabric (metaphysics) of computation that no amount of money, time, or effort can change.
“the principles that he lays out for how general computing devices are possible, which we can go into, it turns out to be exactly the principles that hackers exploit when they hack a computer”
- Building on Gödel's insight that any code can be converted into a unique number, Turing showed that code and data—seemingly opposite categories (active vs passive)—can be represented by the same symbols (binary ones and zeros), so a computer needs to understand only one language and need not be rebuilt for every program.
“you can always take code and turn it into a number, and a unique number. And so you could have code and data represented by numbers, by the same symbols.”
- John von Neumann discovered that biological systems are resilient because they are self-replicating and can rebuild their parts, and his work on how self-replication is possible became the basis for how computer viruses and worms are possible—so the very things that make life possible also make malware possible.
“this idea of, like... Von Neumann has, how can... How is self replication possible? Becomes the basis for how computer viruses and computer worms are possible.”
- Because code and data can be represented by the same symbols, a hacker can send code when the computer or user is expecting data—this is one of the two major ways hacking occurs.
“if they can be represented by the same symbols, when the computer or the user is expecting data, the hacker can send code. And that is the other major way in which hacking occurs.”
Upcode—norms, rules, and incentives—shapes security outcomes more than downcode technology does.
- Upcode should be the focus because it shapes downcode (giving people incentives to produce certain types of code) and because downcode uses data produced by upcode—so no matter how good your operating system or cryptography is, it is undermined if corporate or institutional policies hand over credentials or set bad defaults.
“it doesn't matter how good your operating system is. It doesn't matter how good your cryptography is. None of that matters, is if you go to human resources or IT of your company... and you ask them for credentials to somebody else's account and they give it to you.”
- The Mirai Botnet exploited the internet of things by using default passwords (like 'admin' or '123') that shipped with insecure devices such as cameras and toasters—passwords the attackers simply found in publicly available PDF manuals via Google—to build an enormous botnet capable of taking down large parts of the internet, illustrating that good technology is worthless if surrounding policies are bad.
“if there are these... All these default passwords, default passwords which they just found through Google, 'cause they're in the manuals... they build a worm-like... botnet that exploits these default passwords”
- Across all these hacks, when everything bottoms out it is the human being behind the keyboard who is the vulnerability, which is why a law professor (an 'upcoder') focused on changing rules and incentives is well positioned to improve cybersecurity by ensuring people produce good code or aren't fooled by bad code.
“When all bottoms out, it's like human beings behind the keyboard.”
- Downcode is all the code below your fingertips (operating systems, applications, network protocols, firmware), while upcode is all the norms and rules above your fingertips (psychology, social norms, legal norms, professional ethics, terms of service, employment contracts); standard cybersecurity thinking focuses almost purely on downcode, but addressing upcode—the incentives shaping how code is built and used—is the better approach.
“it doesn't matter how good your operating system is. It doesn't matter how good your cryptography is. None of that matters, is if you go to human resources or IT of your company... and you ask them for credentials to somebody else's account and they give it to you.”
- Paris Hilton's 2005 'phone hack' was actually a hack of the cloud, accomplished by 16-year-old Cameron LaCroix who social-engineered a T-Mobile manager into handing over system credentials and then made his browser impersonate a Sidekick device to access her account data—demonstrating that bad corporate authentication and policy, not phone theft, were the vulnerability.
“Her cell phone is not hacked, it is the cloud that is hacked and it is hacked by a 16-year-old boy known as Cameron LaCroix who figures out how to exploit web interfaces in order to gain access to data on cell phones.”
Humans are the persistent vulnerability in every breach, from phishing to social engineering to poor defaults.
- For ordinary people, hackers don't care about you or your personal data—they run a high-volume business scanning the internet and sending phishing emails to find reckless, easy targets they can pull into a botnet or whose banking and credit card information they can exfiltrate or ransom; if you simply avoid being reckless (don't click unknown links, update your computer), you will likely be fine because you only need to be faster than the next person.
“most people... Hackers don't care about you. They don't even care about your data so much... What they want is to make money. And it's a high volume business.”
- Despite the Russian state's massive resources, the crucial step in breaching the Clinton campaign was John Podesta's IT consultant writing 'This is a legitimate email' (apparently meaning to write 'not a legitimate email') in response to a Fancy Bear phishing email, leading Podesta to hand his Gmail credentials to Russian military intelligence—a serious human vulnerability with major political consequences.
“John Podesta sends his email to IT saying, 'Hey is this legitimate?' And the guy writes back, 'This is a legitimate email.' And he claims he meant to say, 'This is not a legitimate email.'”
- Because computation is a physical process, hackers can read off information from changes in the physical system—called a side-channel attack—such as using a smartphone's accelerometer to detect which numbers were pressed when entering a passcode.
“there are all these cool things where hackers can discover your passcode because you have an accelerometer in your smartphone, which can tell which numbers you've pressed”
- Hacking is best defined as an activity that attempts to defeat a security control (such as a password requirement); merely reading someone's email on an unlocked computer is not hacking because no control designed to stop you was defeated, whereas using fraud or deception to work around such a control is hacking.
“when I use the term hacking, I use it as an activity that attempts to defeat a security control.”
State-level hacking follows international norms of quiet espionage until those norms are publicly violated.
- The norm is that states may invade each other's digital systems to collect national security information but are supposed to keep it quiet; Cozy Bear followed this by quietly collecting intelligence for analysis, but Fancy Bear broke the norm with a massive public dump of information, transforming espionage into something some have called cyber war—and this single change in upcode caused the disaster.
“Fancy Bear takes this information and does something which had not been done before, which is this massive dump of information, which takes something from espionage and turns it more into something that some people have been tempted to say, is cyber war.”
- When hacking is done to collect national security information it is espionage, which is legal under international law—which is why every state hacks every other state—and this upcode reality explains why the FBI was slow to act and unalarmed about Russia being in the DNC network ('dog bites man'), since state-on-state spying is normal and expected.
“when hacking is done for the purposes of collecting natural security information, it is known as espionage, spying. And the spying is legal under international law, that is why every state hacks every other state.”
- In a world of pervasive social media, it is unbelievably difficult to be anonymous, so intelligence agencies struggle to field spies—open-source intelligence groups like Bellingcat can mine Russian social media to identify operatives, and Fancy Bear members had such poor operational security that they registered their cars to the unit's address.
“in a world where everyone is connected in a world of social media it's unbelievably difficult to be anonymous.”
Young skilled hackers are socially motivated offenders best diverted through legitimate competitive venues and employment.
- Because hackers are bored, skilled young men driven by fun and social recognition, a promising remedy—implemented in the UK and Netherlands and being caught up to by the US—is to create legitimate hacking venues like capture-the-flag competitions and mentoring to divert young offenders into the legitimate cybersecurity industry rather than into black-hat activity.
“to try to create legitimate hacking venues. And to try to divert young offenders into the legitimate cybersecurity industry, as opposed to the, the black hat activities.”
- Much of cybercrime is a response to underemployment in less developed tech economies—Bulgaria became a virus powerhouse in the 1980s because it had many underemployed young men with excellent engineering educations and good coding skills but no jobs, so they made viruses for fun.
“so much of cyber crime is a response to underemployment in less developed tech economies.”
- Contrary to the loner stereotype of neurodivergent freaks, hackers are largely ordinary, socially motivated people who want clout—they want their online peers to think they are excellent hackers—and this social drive is so strong that coders will release their code to get peer recognition even at the risk of getting caught.
“hackers are not loners. The picture that they have is that they are freaks... But that's just kind of not true... they want clout. They want their peers online to think that they are excellent hackers.”
- Instead of jailing the three young men behind the Mirai Botnet, the court ordered 2,500 hours of community service helping the FBI catch other malware purveyors—a 'catch me if you can, patch me if you can' model that Shapiro considers an excellent way to deal with skilled young offenders.
“instead of putting them in jail, the court orders them for 2500 hours of community service, where they are helping the FBI catch a lot of these malware purveyors.”
Malware variants exploit fundamental mechanisms like self-replication and code-data equivalence.
- Hackers routinely lie, often taking the same malware, renaming it with a different anime name, and presenting it as 'new improved' or their own creation, which is why many variants of the Mirai Botnet (such as Satori) still circulate today.
“hackers lie. I hate to break it to you... they'll often take the same malware and they'll rename it with a different anime name. And then say it's the new improved malware.”
- Viruses and worms both use trickery but differ in two main ways: who executes them (viruses trick users into clicking/executing while worms are autonomous and trick operating systems or network protocols) and how they spread (worms spread through networks, making them dangerous on the internet, while viruses are local and do not spread via networks).
“One is who executes them, is it the user... Does the user need to click on the thing? And whereas a worm can... Is autonomous can operate by itself. And the second is, how does it spread?”
- Dark Avenger created a polymorphic virus engine that introduced mutations so every copied virus had a new 'genetic' signature, which defeated signature-based antivirus software that worked by scanning programs for known viral signatures.
“he figures out how to create a polymorphic virus engine... introducing mutations so that every new virus that gets spread has a new genetic signature... which defeated the antiviral software.”
High-value targets face persistent active threats; ordinary users need only basic precautions and relative caution.
- High-value targets—journalists, politicians, C-suite executives, human rights activists, and government officials—must assume they are being actively targeted and should seek professional security help, in contrast to ordinary users for whom basic precautions suffice.
“High value targets, which I'm including journalism, politicians, people in the C-suite... human rights activists, people like that, government official. They are high value targets. They must assume that they are being targeted.”
- After Ronald Reagan saw the film WarGames at the White House and grew alarmed about whether a teenager could really start World War III by connecting to a NORAD computer, the first iteration of the Computer Fraud and Abuse Act—the first federal law against hacking—was passed the following year.
“Ronald Reagan sees this movie at the White House and then freaks out... And we get the next year, the first iteration of the Computer Fraud and Abuse Act, the first federal law against hacking”
- The 1988 Morris worm was effectively the first time most Americans had ever heard the word 'internet'—newspapers writing it up did not even know how to refer to it (the internet, internet, capitalized N, etc.)—so people first learned the internet existed when it crashed.
“the first time people know that the internet exists, basically is when it crashes.”
- A standard pattern in cybersecurity history is that those who purvey the cures are often the same ones purveying the diseases—offering protection services for problems they themselves create—as in Paras Jha's DDoS attacks on Rutgers to force them to use his DDoS-mitigation firm, a classic racketeering 'offer you can't refuse.'
“the ones who purvey the cures are often the ones who are also purveying the diseases... So you offer protection services for problems that you yourself create.”
- Robert Morris Jr.'s 1988 worm exploited multiple vulnerabilities in Unix 4.2 BSD and, though intended only as an experiment to measure the size of the internet, was so effective at reinfecting computers repeatedly that they crashed from being too busy copying and distributing the worm rather than from any intent to crash them.
“the worm was so effective at infecting computers that it reinfected those computers over and over again. And so these computers crashed not because he was trying to crash them, but because they were so busy copying worms and distributing them.”